Anthropic chief Amodei urges US and China to agree on AI controls
Anthropic CEO Dario Amodei called on the United States and China to combine restraint in the development of artificial intelligence with agreements on mutual risk limitations. In an interview with the US television network CBS, he said that strategic rivalry with China is the main obstacle to slowing technological development, Anadolu Agency reports.
US-China agreements
Amodei proposed restricting supplies of advanced semiconductor chips to “authoritarian” states in order to gain time for negotiations with Beijing. In his view, the United States and China should try to establish mutual restrictions on the dangerous use of AI systems.
He compared such negotiations to nuclear arms control agreements between the United States and the USSR. According to the Anthropic chief, even adversaries that had thousands of missiles recognized the need for talks to reduce the scale of potential destruction. At the same time, Amodei stressed that the Cold War itself should not be considered a model to follow.
Ahead of the planned bilateral meeting between US President Donald Trump and Chinese President Xi Jinping, he urged the sides to conclude specific agreements to reduce catastrophic risks. Among the priority steps, Amodei named extending the provisions of the Biological Weapons Convention to AI systems, so that states would not use them to create pathogens or release models that bioterrorists could exploit. He emphasized that the key issue in such agreements is verifying their implementation.
More current news is available on the UA.News Telegram channel Telegram.
External oversight of laboratories
Amodei also proposed involving external evaluators from non-profit organizations or government bodies directly in laboratories during model training. Such inspectors would verify whether developers comply with their stated protocols, similarly to auditors in food safety.
According to the agency, the idea received preliminary support from OpenAI’s Sam Altman and xAI’s Elon Musk. Amodei believes that companies should jointly determine standards for releasing new systems with government authorities.
The Anthropic CEO opposed a legislative ban on advanced AI models because, in his view, it could halt research capable of leading to medical breakthroughs and weaken the country’s technological position. He also warned that an emergency shutdown is not sufficient protection: a powerful system, he said, may circumvent attempts to stop it. Instead, Amodei supported a multilayered protection system with several complementary barriers.
